Data in this report are obtained from the Centers for Disease Control and Prevention’s Behavioral Risk Factor Surveillance System (BRFSS) and National Health Interview Survey (NHIS), and the Substance Abuse and Mental Health Services Administration’s National Survey on Drug Use and Health (NSDUH).
Data are from 2011-2018. To ensure adequate sample size for the number of people who have served, two years of data were combined into four time periods: 2011-2012, 2013-2014, 2015-2016 and 2017-2018. Baseline refers to the first set of data years (2011-2012). Prior refers to the last edition published before the current release, which used 2015-2016 data. Current refers to this year’s report, which uses the most recent data available, 2017-2018.
Data were weighted according to each survey’s weighting methodology to correct for selection bias and ensure representative samples by demographic variables. To reflect the differing age distribution of those who have and have not served, data were age-adjusted to the 2000 U.S. Standard Population. Weighted and age-adjusted point estimates were calculated and are included in this report for those who have and have not served, overall and by gender, age, race/ethnicity, income and education.
Subpopulation categories were reported consistently across all data sources where possible, though in some instances, categories were not comparable across surveys. Age categories for all indicators were reported in accordance with NSDUH age ranges, as NSDUH reports respondent ages only by category and not as a discrete value. NHIS does not report a Hawaiian/Pacific Islander race/ethnic group. Cutoff points for the lowest household income category also differed across surveys. For purposes of general alignment, the following categories were selected for the lowest annual household income cutoff in each survey: $25,000 for BRFSS, $30,000 for NSDUH and $35,000 for NHIS.
Data were suppressed according to the guidance provided by each data source. For BRFSS and NHIS, weighted prevalence estimates were suppressed if the relative standard error was greater than 30% or if the unweighted sample size was less than 50. For NSDUH, weighted prevalence estimates were suppressed if the estimated prevalence rate was less than 0.00005, greater than 0.999995 or if the unweighted sample size was less than 100. Additionally, data were suppressed if the relative standard error was greater than 0.175 (according to suggested NSDUH methodology) and/or if the effective sample size was less than 68.
Given the large annual sample sizes in the analyzed datasets and the pooling of multiple years of data to produce estimates, the numbers presented on those who have served are backed by adequate statistical power. Further, the sampling designs of these surveys ensure representation by multiple demographic variables.
However, there are limitations to interpreting data on those who have served. For example, each of the three sources of data analyzed for this report asks different questions about military service. Since 2011, BRFSS has asked only whether the respondent has served on active duty in the U.S. Armed Forces. By comparison, NSDUH asks whether respondents have ever been in the U.S. Armed Forces and excludes any who are currently on active duty. NHIS asks if the respondent has ever served in the U.S. Armed Forces, Military Reserves or National Guard and excludes those on active duty. As such, BRFSS data in this report do not distinguish between those currently serving and those who have been discharged, while NSDUH and NHIS data exclude those on active duty but include those who currently or in the past have served in the Military Reserves or National Guard without being activated. For the time period analyzed, none of the surveys allows analysis by the nature of discharges, involvement in active combat or the era in which one served. Thus, changes over time could be influenced by cohort effects and may confound the interpretation of age-specific results and comparisons.
Additionally, samples of those who have served and not served may be different from one another in demographic composition, for example citizenship status. Such differences may contribute to observed differences in results between the groups. Caution should also be taken when interpreting data on specific health measures. Of note, many health outcome measures indicate whether respondents have been told by a health care professional that they have a disease, excluding those who may not have received a diagnosis or not have sought or obtained treatment.